IJCA Vol 4 i1 2025 webmag - Flipbook - Page 19
2025 | Volume 4, Issue 1
production and assembly areas
within the BMW Group. Depending
on the delivery condition—such
as just-in-time (JIT), just-insequence (JIS), and stock
(LB)—as well as the complexity
of installation, accessibility,
and module assignment (i.e.,
assigning components to specific
assemblies), there are partial
distinctions between components
(BMW Group, 2021; 2023a).
Existing classification concepts
in the literature have primarily
focused on categorizing different
materials (Roy et al., 1995;
Dixon et al., 2006; Saralajew,
2019; Altenbach et al., 2004). In
the context of component IDs,
however, additional factors—
such as installation complexity,
accessibility, module assignment,
and delivery condition—must also
be taken into account. Therefore, a
specific component classification
is required for application in the
CID (BMW Group, 2021).
2.2 PoCs of Optoelectronic and
Transmitter-Receiver Systems
This section examines previously
conducted proof-of-concept (PoC)
studies involving optoelectronic
and transmitter-receiver systems
Findings from Existing PoCs on
Transmitter-Receiver
Systems Bauer (2019) explains
that information exchange in
electromagnetic transmitterreceiver systems occurs through
signal transmission. The
transmitter generates signals
that are transmitted either via
electromagnetic waves. These
signals are captured by antennas
attached to the relevant objects
and forwarded to the receiver
(Hesse et al., 2014). The receiver
then interprets the signals to
reconstruct the transmitted
information. Common examples
include radio frequency
identification (RFID) (Kern, 2007;
Jodin et al., 2012) and near field
communication (NFC) (Shidaganti,
2021; Want, 2011).
Previous investigations identified
RFID as a suitable transmitterreceiver technology for detecting
component IDs (homologation
labels). Accordingly, this analysis
focuses on RFID.
Within the scope of PoCs,
factors influencing RSSI values
on metallic surfaces have been
further examined (Curran et al.,
2013). It was found that the RSSI
values decrease when RFID tags
are attached to metallic surfaces.
Silva et al. (2018) developed
a cost-effective concept for
determining the operational
efficiency of UHF RFID systems in
the aviation industry and identified
electromagnetic interference as
a significant influencing factor.
Jeevagan et al. (2014) discussed
challenges in mounting RFID tags
on metal surfaces of vehicles and
proposed a theoretical model for
RFID use in vehicle collisions.
Existing research shows that
analyzing minimum activation
power provides important insights
for specific materials (Curran et al.,
2013; Silva et al., 2018). However,
direct comparisons across
different materials, particularly
for CoP components, are lacking.
Many PoCs were not tested in
actual production environments
(Jeevagan et al., 2014; Tuan, 2012).
Therefore, further research is
needed to compare different RFID
tags under both ideal and realworld production conditions.
Findings from Existing PoCs on
Optoelectronic Systems
Böhmer, Ehrhardt, and Oberschelp
(2010) explain that optoelectronic
systems identify objects based
19
on contours or labels such as
colors, reflective markers, fonts,
symbols, or barcodes. Detection
is performed using optoelectronic
sensors—such as laser scanners or
cameras—that illuminate the object
with an external light source and
capture the reflected light (Hesse
et al., 2014).
Previous studies have identified
optical character recognition
(OCR) as a suitable detection
technology for optoelectronic
systems to detect component
IDs (homologation labels), among
others. This section therefore
focuses on OCR.
OCR is now a subfield of computer
vision that extends beyond
detecting content in printed
documents (Chaudhuri, 2017).
One practical application is the
automated extraction of printed
information on component labels.
The BMW Group currently uses
three OCR models: Tesseract,
EasyOCR, and PaddleOCR (BMW
Group, 2023b). These models differ
in algorithm type, neural network
architecture, and layer connectivity.
Tesseract is widely used and has
industrial applications (Ramadan S
et al., 2023; Brisinello et al., 2017;
Bugayong et al., 2022). EasyOCR
is employed in tasks such as
automated license plate detection
(Sainui et al., 2024; Salsabila et
al., 2024; Sarhan et al., 2024).
PaddleOCR is another OCR tool
that offers strong performance in
extracting printed label content
(Bagaria et al., 2024).
Application of GPT-4v and Insights
from Existing PoCs
Intelligent Character Recognition
(ICR) and OCR are both used for
text detection. The main difference
lies in the type of text each can
recognize: OCR detects printed or
machine-written text (Schmalz,
